intro
- cross-validated cosine similarity on sustained regressor (task versus rest)
- cross-task: leave one task out
Purpose
quick look: all correlations

Parcel-wise stats

- values displayed are group-mean distance d
- images are thesholded at d > 0 and p < 0.05 FDR (sign-rank test) corrected across all 400 parcels
Axcpt

Cuedts

Stern

Stroop

all tasks: conjunction analysis
- parcels with pFDR < 0.05 in all tasks

all tasks: fixed-effect analysis (average across tasks)
- parcels with pFDR < 0.05 in all tasks

DMCC 35 versus null 35
## # A tibble: 8 x 5
## # Groups: task [4]
## task parcelset p statistic b
## <fct> <chr> <dbl> <dbl> <dbl>
## 1 Axcpt dmcc35 0.0000101 5.33 0.219
## 2 Axcpt null35 0.0000597 4.69 0.102
## 3 Cuedts dmcc35 0.000252 4.17 0.186
## 4 Cuedts null35 0.00000238 5.85 0.0909
## 5 Stern dmcc35 0.000134 4.40 0.199
## 6 Stern null35 0.0118 2.69 0.0663
## 7 Stroop dmcc35 0.0000154 5.18 0.210
## 8 Stroop null35 0.000542 3.89 0.0866
## # A tibble: 4 x 4
## task p statistic b
## * <fct> <dbl> <dbl> <dbl>
## 1 Axcpt 0.00227 3.35 0.117
## 2 Cuedts 0.0187 2.49 0.0955
## 3 Stern 0.000457 3.95 0.133
## 4 Stroop 0.000927 3.69 0.123

relationship with hilo contrast

## # A tibble: 4 x 4
## task r_hilo_mu r_taskrest_mu r_taskrest_hilo
## * <chr> <dbl> <dbl> <dbl>
## 1 Axcpt 0.390 0.0764 0.120
## 2 Cuedts 0.0940 0.237 -0.0225
## 3 Stern 0.324 0.143 0.123
## 4 Stroop 0.382 0.299 0.218
- above shows correlation across parcels. “mu” is univariate hilo contrast.